arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 2510.16751cs.CV
收藏

视觉自回归模型在推理时间缩放上超越扩散模型

Visual Autoregressive Models Beat Diffusion Models on Inference Time Scaling

Technical University of Denmark(丹麦技术大学);Pioneer Center for AI(先锋人工智能中心)

Erik Riise, Mehmet Onurcan Kaya, Dim P. Papadopoulos

AI总结:

本文证明视觉自回归模型凭借离散标记空间和束搜索,在推理时缩放上超越扩散模型,使2B参数模型在多个基准上胜过12B参数扩散模型,凸显架构对推理优化的关键作用。

摘要:

虽然通过搜索进行推理时间缩放已经彻底改变了大型语言模型,但将这些收益转化为图像生成却被证明是困难的。最近将搜索策略应用于连续扩散模型的尝试显示出有限的益处,简单的随机采样往往表现最佳。我们证明了视觉自回归模型的离散、顺序特性能够实现有效的图像生成搜索。我们展示了束搜索显著改善了文本到图像的生成,使得一个2B参数的自回归模型在多个基准上超越了12B参数的扩散模型。系统性消融研究表明,这一优势来自于离散的标记空间,它允许早期剪枝和计算重用,而我们的验证器分析则突出了速度与推理能力之间的权衡。这些发现表明,模型架构,而不仅仅是规模,对于视觉生成中的推理时间优化至关重要。

英文摘要:

While inference-time scaling through search has revolutionized Large Language Models, translating these gains to image generation has proven difficult. Recent attempts to apply search strategies to continuous diffusion models show limited benefits, with simple random sampling often performing best. We demonstrate that the discrete, sequential nature of visual autoregressive models enables effective search for image generation. We show that beam search substantially improves text-to-image generation, enabling a 2B parameter autoregressive model to outperform a 12B parameter diffusion model across benchmarks. Systematic ablations show that this advantage comes from the discrete token space, which allows early pruning and computational reuse, and our verifier analysis highlights trade-offs between speed and reasoning capability. These findings suggest that model architecture, not just scale, is critical for inference-time optimization in visual generation.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑